Anthropic

Claude Sonnet 5

Released Jun 30, 2026 · knowledge cutoff 2026-01

Status: Frontier
Location: United States
Modality: Multimodal
Context window: 1M
Max output: 128K
Speed (max effort): 83 tok/s · 202s to first answer token
Price ($/MTok in / out): $2 / $10
Cost per task (max effort): $1.53
Open weights: No

Benchmarks5 of 8 reported

GPQA Diamond

91.1%

SWE-bench Verified

85.2%

Terminal-Bench 2.1

80.4%

Humanity's Last Exam

43.2%

LMArena Elo

1461

MMLU-Pro

—

AIME (latest)

—

ARC-AGI-2

—

Strengths & weaknesses

Strengths

Fast — responses land in seconds even with thinking enabled
Reads intent unusually well; it judges what you meant rather than what you literally typed
Steady enough on long agentic coding runs to be left running unattended

Weaknesses

The most verbose Claude yet — it writes long even when told to be brief, and that length is where the cost goes
Personality divides users; some find it blunt or contrarian outside coding work
Deliberately weak at offensive security tasks

Introductory pricing $2/$10 per MTok through Aug 31 2026, rising to standard $3/$15 on Sept 1. HLE 43.2% no tools / 57.4% with tools per Anthropic's system card. LMArena Elo 1461 is the 'Claude Sonnet 5 High' text-arena listing (arena.ai). SWE-bench Verified 85.2% is widely cited by third-party trackers (OpenRouter, Requesty, EdenAI), but Anthropic's own system card reports SWE-bench Pro (63.2%) instead — no official Verified figure confirmed. GPQA Diamond 91.1% is Artificial Analysis' own measurement (its gpqa-diamond leaderboard, relayed by BenchLM), not an Anthropic figure — a 96.2% number circulating on some trackers is unsourced and was rejected. ARC-AGI-2 and AIME remain unpublished: Anthropic did not report AIME for Sonnet 5 and ARC Prize has not tested it.

For developers

API model strings

anthropicclaude-sonnet-5
aws-bedrockanthropic.claude-sonnet-5

Licence

Proprietary — weights not released

Retirement

No retirement announced

Lineage

Replaces Claude Sonnet 4.6